|Publication number||US6687563 B1|
|Application number||US 10/355,998|
|Publication date||Feb 3, 2004|
|Filing date||Jan 31, 2003|
|Priority date||Jan 31, 2003|
|Publication number||10355998, 355998, US 6687563 B1, US 6687563B1, US-B1-6687563, US6687563 B1, US6687563B1|
|Inventors||Ming Wang, Jim Chue, Ray Shih, Eric Chang, Jason Liu|
|Original Assignee||Taiwan Semiconductor Manufacturing Company|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (14), Referenced by (30), Classifications (11), Legal Events (4)|
|External Links: USPTO, USPTO Assignment, Espacenet|
1. Field of Invention
This invention relates generally to a system and method of planning, scheduling, and dispatching during wafer production in a Fab and, more particularly, to an automated decision engine that integrates short-term (dispatching) and long-term (scheduling) production plans for optimum dispatching and scheduling of wafer build throughout the wafer's movement in the Fab.
2. Description of Related Art
In semiconductor manufacturing Fabs costing billions of dollars efficient use of tools is essential to controlling costs and maximizing throughput of the product. During the building of wafers, the wafer box or Front Open Unified Pot (FOUP) travels from position to position to be placed into the next required tool for its next processing to take place. A Fab planner does long-term production planning and calculates an optimal path using a scheduling tool. Separately, a Fab supervisor planner is doing short-term production planning to balance and filter WIP (Work in Progress) with a dispatching tool to set up an optimal path based on what is actually happening currently on the Fab floor. Setting up different rule for different product areas is usually done using a dispatching tool. As the planning is done differently and at different times, the plans are frequently inconsistent with effort sometimes being duplicated. Integrating of the two plans can be very difficult. Furthermore, individual usage of different reports by both the planner and supervisor can result in conflict between reports, subjectiveness, and increased costs due to software efforts needed to create support for this.
Most of the dispatching system has many exception rules, making it inefficient and very difficult to maintain. Furthermore, most dispatching systems do not consider the current position of the FOUP, so the optimal decision cannot be reached on the next move to make in a dynamic Fab like the new 300 mm Fab. In the more complex 300 mm Fabs, an Automatic Machine Handling System (AMHS) must choose the FOUP to be transferred by the dispatch system. As the AMHS is not rule based, it is difficult to modify rules or easily set up rules for each equipment type.
An additional difficulty comes from the fact that the scheduling tool has a performance issue, making it difficult for the supervisor to get real-time data from it. The bottom line is that current full-automation systems such as the 300 mm Fab with an AMHS can optimize MES logical data but can not optimize stocker utilization or avoid traffic jams caused by the AMFHS system.
A new method and system is needed to resolve these problems and optimize the FOUP travel through the Fab. This invention solves these problems and provides additional benefits.
In U.S. Pat. No. 5,880,960 (Lin et al.) a method to improve WIP balance in the manufacturing line is discussed. In U.S. Pat. No. 5,751,580 (Chi) a fuzzy logic method and system for adjusting of priority rating of work is discussed. In U.S. Pat. No. 5,623,413 (Matheson et al.) a scheduling system and method for moving objects through a multipath system is discussed. In U.S. Pat. No. 5,612,886 (Weng) a method and system for the dynamic dispatching of work in progress is discussed. In U.S. Pat. No. 5,548,518 (Dietrich et al.) an allocation method for generating a production schedule is discussed. In U.S. Pat. No. 5,546,326 (Tai et al.) a dynamic dispatching method is discussed. In U.S. Pat. No. 5,446,671 (Weaver et al.) a look-ahead method for maintaining optimum queued quantities is discussed. In U.S. Pat. No. 5,444,632 (Kline et al.) an apparatus and method for controlling and scheduling processing machines is discussed. In U.S. Pat. No. 5,229,948 (Wei et al.) a method of optimizing a serial manufacturing system is discussed.
This invention's overall objective is to provide an integrated method and system for the automation of a semiconductor Fab for production decision support. More specifically, the objective is to provide for a software decision engine to supervise optimal throughput of product into processing equipment. It is also an objective that the decision engine be rule based and that it define and integrate with a cooperative model between dispatch and schedule functions. It is another objective to have other Fab systems also integrate with the decision engine to interact as a unified whole system.
Additionally, it is an objective to provide a decision evaluation formula to integrate the results of system analysis tools for production decision support. It is an objective to have a sub-formula of the decision evaluation formula to resolve dynamic constraint issues. It is also an objective for the method and system to provide for an automatic learning methodology.
These objectives are achieved by the method and system of this invention. The use of this method and system in a Fab system provides cost savings by efficient automation and by optimizing the throughput of product traveling from tool to tool. It reduces bottle necks and over and under supply to tools by allowing each tool equipment control program to help in the movement of product between tools.
This invention will be described with reference to the accompanying drawings, wherein:
FIG. 1 is an illustration diagram comparing the new decision engine and prior art general system.
FIG. 2 is a flow diagram of the prior art process.
FIG. 3 is a block diagram of the new system architecture.
FIG. 4 is a flow diagram showing the relationship between the OJS, equipment, and AMHS in full-automation processing.
FIG. 5 a flow diagram of the method of the system.
In complex semiconductor manufacturing Fabs today, optimum scheduling and dispatching for product as it is manufactured plays a key role in cost-effective production. If tools stand by idle or have too much work waiting for them, the cost of production goes up as the tools are not being used optimally. Complex automated software systems have been developed to aid tool usage. These automated systems were developed in piecemeal fashion as the need arose to control or aid specific areas of the manufacturing process.
Fab planners that have the responsibility to control all the tools in the Fab use a scheduling function for long-term planning to best get a Front Open Unified Pot (FOUP) or wafer box through the various tools needed to create wafers. Each FOUP is one of several in production at any given time, all vying concurrently for a limited number of tools. Supervisors do short-term planning using a dispatching function. The goal of both planner and supervisor is to keep all tools busy and at maximum production by getting the FOUP to the correct tool at the correct time following an optimum path. This current schedule and dispatch system for balancing and filtering of Work in Progress (WIP) is based on individual usage of reports. This has the disadvantages of conflict between reports (ex. One report shows the FOUP should go to equipment A, while another shows that equipment B is idle.), subjectivity, and the high maintenance costs due to the many programs needed to support it. This plus the many complexities that occur on the Fab floor make it difficult to have consistency between scheduling and dispatching and to verify if consistency is not achieved. Effort is sometimes being duplicated and much maintenance is required. An automated method and system to integrate both is needed to achieve optimum WIP balance and the most wafers per hour.
In FIG. 1 an illustration diagram compares the prior art general system with the new decision engine and reveals that under the prior art general system 104 rule revision necessitates recompiling and temporary system shut down. A new method and system such as the decision engine model 102 is needed to prevent system shut down for rule revision.
Adding to the complexity is that newer Fabs that control 300 mm wafers are now run overall by an Automatic Machine Handling System (AMHS) that effectively interfaces with the dispatching system. When dispatching no longer is correct from the coded rules, the AMHS does not know it and causes utilization to fall by overstocking or no stocking at tools or by sending a FOUP on a longer path to its next tool. FIG. 2 is a flow diagram that illustrates this problem. The AMHS 202 contains the FOUP Schedule, the Dispatcher, and other AMHS applications. The AMHS controls the operations on the Fab Floor 204. Tool group A is a set of tools 1A and 2A that perform the same process. Similarly, tool group B has tools 1B, 2B, and 3B, and tool group C has tools 1C and 2C. Transportation of FOUP is via transport system paths 208 that frequently are automatic overhead movers.
If the FOUP Schedule 209 is developed to use tools 1A, 2B, and 2C, and tool 2B goes off line while the FOUP is being operated on by tool IA, it is up to the Dispatcher 210 to correct the processing to another tool. As work continually is added to the Fab, the Dispatcher can be overwhelmed and work can start building up at tools 1B and 3B. If additionally tool 1C goes off line and 2B comes back online, the Dispatcher 210 must again try to correct the situation. Without current input into the Dispatcher, delays and overstocking can result before the FOUP is Done 212 with its processing. Transport condition, dynamic constraints, and special conditions are not easily factored in under the current process.
Thus, a new method and system is needed to prevent these problems of scheduling and dispatching of the FOUP's. This invention provides this solution and other advantages.
The new method provides a rule-based software decision engine called the Operation Job Supervisor system (OJS) to supervise optimal throughput of product into processing equipment. This decision engine integrates with other Fab systems (such as AMHS and MES (Manufacturing Equipment System) and a cooperative model between dispatch and schedule functions that provides consistency. FIG. 3 shows a block diagram of the systems architecture. The Schedule 302 tool gives input to the Dispatch 304 tool. The Dispatch tool is now integrated with the Decision Engine 310 through a Decision Integrator 306 which also integrates with the AMHS. This is the same AMHS that in the prior art included the material handling system and its hardware. The Decision engine 310 includes the Operation Job Supervisor system and the transportation interface of the MES, the Move Agent, the CIM (Computer Integrated Manufacturing) database, and the Job Agent. The Job Agent communicates with the Equipment Controllers 312 at each individual tool's host controller. As an added control feature, a Dispatch or Schedule Viewer 314 is available to check operations in realtime on a networked PDA or other device.
Looking at the OJS system in detail, it handles three functions for control of the FOUP. These functions are equipment asking FOUP or wafer box to move in (MIR), equipment asking FOUP or wafer box to move out (MOR), and having FOUP or wafer box go into stocker (Stock-in). The equipment host program executes calls to the decision engine when FOUP is transporting or static and the equipment host program asks the decision engine to adjust decision (Re-MIR).
The OJS is a fully automated decision engine that is integrated with the schedule and dispatch systems. Every MIR command will use dispatch results, and every MOR command will use schedule results. These commands are issued by the actual equipment host program of the tools being used. Thus, schedule and dispatch systems are now integrated with the tools. When a FOUP process is completed, MOR is called to get the next equipment and a call is made to the equipment's MIR. This tool-to-tool transportation will include schedule results (by MOR) and dispatch results (by MIR). The OJS monitors these commands and transport. With this information it can detect and avoid stocker traffic jams and optimize stocker crane utilization. The OJS can use dispatch and schedule results to support overhead external buffers (OHB), stockers, and special storage. It can pre-fetch FOUP's into the stocker or send them to the OHB directly.
The scheduling function determines the next suitable equipment for a FOUP by a schedule system algorithm and has an added function of calculating a transportation cost table to arrive at routing costs. The dispatch system will decide which sorted Work In Progress (WIP) should next be processed using the transportation cost table from the schedule function and determine the next suitable equipment for a FOUP to achieve maximum benefit using the schedule result.
Additionally, this new method and system combines the dispatch rules with dispatch or scheduling system by a registration and reservation technique. For registration, the OJS delivers the information (ex. lot destination and priority) and current AMHS status to the dispatch system. For reservation, the dispatch system should follow the information that OJS delivers to meet its own optimal criteria but can change it according to the newest situation or special condition. For example, when an equipment A's host port program asks for a MIR (request to load a FOUP as it is currently empty) the MIR calls the dispatch system to get next FOUP from a WIP sorted list. If the schedule system had chosen FOUP B as the next candidate for equipment A, the OJS will pass this information to the dispatch system by registering FOUP B to equipment A. The dispatch system can accept this result or change this decision by looking at the current situation.
A central decision evaluation formula is a core part of the OJS decision engine. Its purpose is to minimize costs. It is a dynamically calculated formula that eventually provides the decision in engine with information from system analysis tools including schedule and dispatching to make an optimal decision on transport. The value “Formula” that results is a number calculated by using values obtained from several real-time automated systems that are normalized by using predefined parameter constants. These constants have been chosen by trial and error to insure that each value is weighted properly and corrects for unit differences. This formula value is calculated for each lot and its relative value is the bases for determining which lot has the highest priority. A simple table with the calculated values for each lot sorted highest to lowest is all that is needed to make the decision on which lot to transport. The lowest calculated value being the highest priority lot, and the first to be dispatched. The next lowest is in turn chosen, keeping in mind that the values are recalculated in real time so that priorities are adjusted as changes are introduced to the manufacturing operation. The formula is:
Formula=c1*S+c 2*D+c 3*T+c 4*C+c 5
S is the scheduling system evaluation result for each equipment.
D is the dispatch system evaluation result for each equipment.
T is the distance to each equipment compared to current position for calculating transportation costs.
C is the operational or reset cost. If FOUP is transferring, C is equal to 1. Otherwise it is zero. c1, c2, c3, c4, c5 are parameter constants.
A dynamic sub-formula that can consider transport condition, constraint issues, and special conditions is given below. Again, as before, the resultant value “D” is a number calculated by using values obtained from several real-time automated systems that are normalized by using predefined parameter constants. These constants have been chosen by trial and error to insure that each value is weighted properly, and corrects for unit differences. This formula value is calculated for each lot as before and its relative value is the bases for determining which lot has the highest priority. A simple table with the calculated values for each lot sorted highest to lowest is again all that is needed to make the decision on which lot to transport. The lowest calculated value being the highest priority lot, and the first to be dispatched. The next lowest is in turn chosen, and values are recalculated in real time so that priorities are adjusted as changes are introduced to the manufacturing operation.
P is the lot priority in the Material Engineering System (MES).
Q is the current time divided by the queue time limit or zero.
M is return this area value if this batch is in maximum batch size.
R is the related resource availability like the reticle or monitor lot.
O is the other constraints such as an MES or special condition manufacturing constraint.
c1, c2, c3, c4, c5 are constants.
The integration of the dispatch and schedule systems and the decision engine in combination with these decision evaluation formulas allows for dynamic, self-adjusting throughput. An automatic learning methodology learns patterns automatically from these formulas in a fashion similar to artificial intelligence when variables are changed to control condition and parameters are changed to control weight for maximum wafers per hour (WPH). Information dynamically transported to the decision engine is a learning source.
FIG. 4 is a flow diagram that shows the relationship between the OJS, equipment hosts, and the AMHS in full automation processing. The first operation shown is that the Equipment Host Program 402 with its tool currently empty issues a MR move-in request calling the dispatch rules to the Decision Engine (OJS, MES) 404. The Decision Engine, based on the formula, scheduling and dispatcher rules, chooses the best FOUP and asks FOUP to move by sending a request to the AMHS 406. The AMHS then sends FOUP to Equipment 408. The Equipment tells the host program when the FOUP arrives. When processing is done by the Equipment, the Equipment Host Program 402 sends a MOR move-out request calling the scheduling rules to the Decision Engine 404 which uses the scheduling rules to help choose the next tool. The Decision Engine then informs another Equipment Host Program 410 that there is a FOUP to be processed. That Equipment Host Program then issues the MIR move-in request and the sequence is started again.
The new method can be seen by looking at the system flow diagram of FIG. 5. The output of the Schedule Tool 502 generates schedule results periodically and places them in the Database 504. The Equipment Host Program 506 issues a MIR move-in request call to the Decision Engine 508. The Decision Engine asks the Dispatcher Tool 510 to re-calculate WIP. Scheduling data is retrieved from the Database 504 and the Dispatcher Tool 510 passes its decision derived using the formulas back to the Decision Engine 508. From that information the Decision Engine chooses a lot from the dispatcher and informs the Equipment Host Program 506. The FOUP is moved to that equipment and processing begins. When processing is completed, the Equipment Host Program issues a MOR move-out request to the Decision Engine 508. The Decision Engine asks the Dispatcher Tool 510 to get next equipment. The Dispatcher Tool queries data including schedule results that are now more current from Database 504. The Dispatcher Tool sends its decision back to the Decision Engine 508 and the next Equipment Host Program 512 is notified and the FOUP is prepared to be sent to the next equipment.
The method of the invention provides advantages over the prior art including that the system provides full automation and will dynamically learn to choose the optimum path to directly transfer the FOUP for its next processing to maximize throughput and reduce cost with less human involvement, the system architecture is easier to maintain, and the integration of dispatch and schedule functions results in greater ability and a more consistent and less fuzzy result. Furthermore, central evaluation formulas dynamically collect information from system analysis tools for improved transport decisions for a reduction of transport time and costs, and operation costs are reduced by considering dynamic constraints and special conditions.
While the invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made without departing from the spirit and scope of the invention.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US5229948||Nov 3, 1990||Jul 20, 1993||Ford Motor Company||Method of optimizing a serial manufacturing system|
|US5444632||Apr 28, 1994||Aug 22, 1995||Texas Instruments Incorporated||Apparatus and method for controlling and scheduling processing machines|
|US5446671||Oct 22, 1993||Aug 29, 1995||Micron Semiconductor, Inc.||Look-ahead method for maintaining optimum queued quantities of in-process parts at a manufacturing bottleneck|
|US5546326||Apr 4, 1995||Aug 13, 1996||Taiwan Semiconductor Manufacturing Company Ltd||Dynamic dispatching rule that uses long term due date and short term queue time to improve delivery performance|
|US5548518||May 31, 1994||Aug 20, 1996||International Business Machines Corporation||Allocation method for generating a production schedule|
|US5612886||May 12, 1995||Mar 18, 1997||Taiwan Semiconductor Manufacturing Company Ltd.||Method and system for dynamic dispatching in semiconductor manufacturing plants|
|US5623413||Sep 1, 1994||Apr 22, 1997||Harris Corporation||Scheduling system and method|
|US5751580||Jul 26, 1996||May 12, 1998||Chartered Semiconductor Manufacturing, Ltd.||Fuzzy logic method and system for adjustment of priority rating of work in process in a production line|
|US5826238 *||Apr 2, 1996||Oct 20, 1998||Taiwan Semiconductor Manufacturing||Daily target generation and machine allocation with priority|
|US5880960||Jan 27, 1997||Mar 9, 1999||Taiwan Semiconductor Manufacturing Company, Ltd.||Method to improve WIP balance in a manufacturing line|
|US6128588 *||Oct 1, 1997||Oct 3, 2000||Sony Corporation||Integrated wafer fab time standard (machine tact) database|
|US6341240 *||Jun 4, 1999||Jan 22, 2002||International Business Machines Corporation||Method of allocating work in capacity planning|
|US6456894 *||Jun 1, 1999||Sep 24, 2002||Applied Materials, Inc.||Semiconductor processing techniques|
|US6564113 *||Jun 15, 2001||May 13, 2003||Advanced Micro Devices, Inc.||Lot start agent that calculates virtual WIP time in a multi-product and multi-bottleneck manufacturing environment|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US7031795 *||Mar 10, 2004||Apr 18, 2006||Taiwan Semiconductor Manufacturing Co., Ltd.||System and method of coinsurance wafer management|
|US7043322 *||Mar 8, 2005||May 9, 2006||International Business Machines Corporation||System and method for batch scheduling of lots based on flow time and number of batches|
|US7054703 *||Sep 9, 2003||May 30, 2006||Taiwan Semiconductor Manufacturing Co., Ltd.||Control system and methods for managing a production operating under time constaraints|
|US7069098 *||Aug 2, 2004||Jun 27, 2006||Advanced Micro Devices, Inc.||Method and system for prioritizing material to clear exception conditions|
|US7206653||Nov 29, 2005||Apr 17, 2007||Taiwan Semiconductor Manufacturing Co., Ltd.||Wafer-based planning methods and systems for batch-based processing tools|
|US7239930||May 24, 2005||Jul 3, 2007||International Business Machines Corporation||Method, system, and computer program product for improved flow of development lots in a manufacturing facility|
|US7305276||Oct 31, 2005||Dec 4, 2007||International Business Machines Corporation||Method, system, and computer program product for controlling the flow of material in a manufacturing facility using an extended zone of control|
|US7426420||Jan 31, 2006||Sep 16, 2008||International Business Machines Corporation||System for dispatching semiconductors lots|
|US7477956 *||Jul 12, 2005||Jan 13, 2009||Applied Materials, Inc.||Methods and apparatus for enhancing electronic device manufacturing throughput|
|US7505828||Jul 14, 2006||Mar 17, 2009||Taiwan Semiconductor Manufacturing Company, Ltd.||Carrier transportation management system and method for internal buffer process tools|
|US7539553 *||Jun 30, 2003||May 26, 2009||Taiwan Semiconductor Manufacturing Co., Ltd.||System and method for automated dispatch and transportation of work-in-process|
|US7664561||Apr 25, 2002||Feb 16, 2010||Taiwan Semiconductor Manufacturing Company, Ltd.||Task queuing methodology for reducing traffic jam and to control transmission priority in an automatic material handling system|
|US8160736 *||Jan 31, 2008||Apr 17, 2012||Globalfoundries Singapore Pte. Ltd.||Methods and apparatus for white space reduction in a production facility|
|US8225496||Aug 29, 2008||Jul 24, 2012||Applied Materials, Inc.||Automated integrated solar cell production line composed of a plurality of automated modules and tools including an autoclave for curing solar devices that have been laminated|
|US8335581 *||Jun 12, 2009||Dec 18, 2012||Globalfoundries Inc.||Flexible job preparation and control|
|US8412368 *||Oct 7, 2010||Apr 2, 2013||Globalfoundries Inc.||Method and apparatus for routing dispatching and routing reticles|
|US8455260||Mar 26, 2010||Jun 4, 2013||Massachusetts Institute Of Technology||Tagged-fragment map assembly|
|US20040267641 *||Jun 30, 2003||Dec 30, 2004||Yao-Hsiung Chang||System and method for automated dispatch and transportation of work-in-process|
|US20050055105 *||Sep 9, 2003||Mar 10, 2005||Kan Wu||Control system and methods for managing a production process operating under time constraints|
|US20050203658 *||Mar 10, 2004||Sep 15, 2005||Jen-Lin Chao||System and method of coinsurance wafer management|
|US20060025879 *||Aug 2, 2004||Feb 2, 2006||Purdy Matthew A||Method and system for prioritizing material to clear exception conditions|
|US20060030966 *||Jul 12, 2005||Feb 9, 2006||Inchen Huang||Methods and apparatus for enhancing electronic device manufacturing throughput|
|US20060271220 *||May 24, 2005||Nov 30, 2006||International Business Machines Corporation||Method, system, and computer program product for improved flow of development lots in a manufacturing facility|
|US20060271223 *||May 27, 2005||Nov 30, 2006||International Business Machines Corporation||Method and system for integrating equipment integration software, equipment events, mes and rules databases|
|US20060282187 *||Jan 31, 2006||Dec 14, 2006||International Business Machines Corporation||System for dispatching semiconductors lots|
|US20070005170 *||Jun 29, 2005||Jan 4, 2007||Thorsten Schedel||Method for the preferred processing of workpieces of highest priority|
|US20100318209 *||Jun 12, 2009||Dec 16, 2010||Globalfoundries Inc.||Flexible job preparation and control|
|US20120089245 *||Oct 7, 2010||Apr 12, 2012||Diwaskar Adhikari||Method and apparatus for routing dispatching and routing reticles|
|CN100419607C||Apr 3, 2006||Sep 17, 2008||台湾积体电路制造股份有限公司||Wafer-based planning methods and systems|
|CN100476661C||May 25, 2005||Apr 8, 2009||先进微装置公司||Method and system for prioritizing material to clear abnormal conditions|
|U.S. Classification||700/121, 700/100, 700/103|
|International Classification||G06F19/00, G05B19/418|
|Cooperative Classification||G05B19/41865, G05B2219/32271, G05B2219/45031, G05B2219/32297, G05B2219/32247|
|Jan 31, 2003||AS||Assignment|
|Jul 6, 2007||FPAY||Fee payment|
Year of fee payment: 4
|Jul 6, 2011||FPAY||Fee payment|
Year of fee payment: 8
|Sep 11, 2015||REMI||Maintenance fee reminder mailed|